# Gemini 3.8 Flash TTS

> Generate expressive multi-speaker speech with Gemini 3.8 Flash TTS. Studio-grade voice fidelity and long-form stability. Try free, pay per use.

## 概览

- **端点**: `POST https://api.muapi.ai/api/v1/gemini-3-8-flash-tts`
- **模型 ID**: `gemini-3-8-flash-tts`
- **类别**: AI 音频
- **变体**: 3.8 Flash
- **系列**: gemini-tts
- **费用**: 每次调用 0.015 credits（部分模型会根据参数动态计算费用）

## API 使用

MuApi 使用**提交后轮询**模式：提交任务，获取 `request_id`，然后轮询预测端点，直到 `status` 为 `completed`。也可以在提交请求中传入 `?webhook=YOUR_URL`，任务完成后接收 POST 回调（无需轮询）。

**身份验证**：在 `x-api-key` 请求头中发送 MuApi 密钥。可在 https://muapi.ai/access-keys 获取。

### 1. 提交任务

```http
POST https://api.muapi.ai/api/v1/gemini-3-8-flash-tts
Content-Type: application/json
x-api-key: YOUR_API_KEY
```

**最小请求（仅必填参数）：**

```json
{
  "speakers": [
    {
      "speaker_id": "Speaker 1",
      "voice_name": "Fenrir",
      "audio_profile": "A stern and weary gatekeeper",
      "accent": "British (RP)",
      "style": "Deadpan",
      "pace": "Natural"
    },
    {
      "speaker_id": "Speaker 2",
      "voice_name": "Puck",
      "audio_profile": "A determined and courageous traveler seeking answers.",
      "accent": "American (Gen)",
      "style": "Empathetic",
      "pace": "Staccato"
    }
  ],
  "dialogue_turns": [
    {
      "speaker_id": "Speaker 1",
      "text": "[shouting] Halt, traveler! The northern pass is sealed by order of the council."
    },
    {
      "speaker_id": "Speaker 2",
      "text": "[determination] I carry a message for the elder. Step aside, or I will force my way through."
    },
    {
      "speaker_id": "Speaker 1",
      "text": "[caution] No one passes. [pensive] The elder is... he's no longer receiving visitors."
    },
    {
      "speaker_id": "Speaker 2",
      "text": "It's too late. [whispers] The shadow... it reached him first. [urgency] You need to leave. [shouting] Now."
    }
  ]
}
```

**完整示例（全部参数）：**

```json
{
  "speakers": [
    {
      "speaker_id": "Speaker 1",
      "voice_name": "Fenrir",
      "audio_profile": "A stern and weary gatekeeper",
      "accent": "British (RP)",
      "style": "Deadpan",
      "pace": "Natural"
    },
    {
      "speaker_id": "Speaker 2",
      "voice_name": "Puck",
      "audio_profile": "A determined and courageous traveler seeking answers.",
      "accent": "American (Gen)",
      "style": "Empathetic",
      "pace": "Staccato"
    }
  ],
  "dialogue_turns": [
    {
      "speaker_id": "Speaker 1",
      "text": "[shouting] Halt, traveler! The northern pass is sealed by order of the council."
    },
    {
      "speaker_id": "Speaker 2",
      "text": "[determination] I carry a message for the elder. Step aside, or I will force my way through."
    },
    {
      "speaker_id": "Speaker 1",
      "text": "[caution] No one passes. [pensive] The elder is... he's no longer receiving visitors."
    },
    {
      "speaker_id": "Speaker 2",
      "text": "It's too late. [whispers] The shadow... it reached him first. [urgency] You need to leave. [shouting] Now."
    }
  ],
  "scene": "",
  "sample_context": "",
  "temperature": 1
}
```

**响应：**

```json
{
  "request_id": "abc123",
  "status": "processing"
}
```

### 2. 轮询结果

```http
GET https://api.muapi.ai/api/v1/predictions/{request_id}/result
x-api-key: YOUR_API_KEY
```

可能的 `status` 值：`queued`、`pending`、`processing`、`completed`、`failed`、`cancelled`。每 2–5 秒轮询一次，直到进入终止状态。状态为 `completed` 时，结果 URL 位于 `outputs` 数组中。

**`completed` 状态的响应示例：**

```json
{
  "id": "abc123",
  "status": "completed",
  "outputs": [
    "https://cdn.muapi.ai/.../output.png"
  ],
  "urls": {
    "get": "https://api.muapi.ai/api/v1/predictions/abc123/result"
  },
  "created_at": "2026-05-08T12:34:56Z",
  "has_nsfw_contents": []
}
```

### cURL

```bash
1. 提交
REQUEST_ID=$(curl -s -X POST https://api.muapi.ai/api/v1/gemini-3-8-flash-tts \
  -H "x-api-key: $MUAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"speakers":[{"speaker_id":"Speaker 1","voice_name":"Fenrir","audio_profile":"A stern and weary gatekeeper","accent":"British (RP)","style":"Deadpan","pace":"Natural"},{"speaker_id":"Speaker 2","voice_name":"Puck","audio_profile":"A determined and courageous traveler seeking answers.","accent":"American (Gen)","style":"Empathetic","pace":"Staccato"}],"dialogue_turns":[{"speaker_id":"Speaker 1","text":"[shouting] Halt, traveler! The northern pass is sealed by order of the council."},{"speaker_id":"Speaker 2","text":"[determination] I carry a message for the elder. Step aside, or I will force my way through."},{"speaker_id":"Speaker 1","text":"[caution] No one passes. [pensive] The elder is... he's no longer receiving visitors."},{"speaker_id":"Speaker 2","text":"It's too late. [whispers] The shadow... it reached him first. [urgency] You need to leave. [shouting] Now."}]}' | jq -r .request_id)

2. 轮询直到完成
while :; do
  RESP=$(curl -s https://api.muapi.ai/api/v1/predictions/$REQUEST_ID/result -H "x-api-key: $MUAPI_API_KEY")
  STATUS=$(echo "$RESP" | jq -r .status)
  [ "$STATUS" = "completed" ] && echo "$RESP" | jq .outputs && break
  [ "$STATUS" = "failed" ] && echo "$RESP" && exit 1
  sleep 3
done
```

### Python

```python
import os, time, requests

API = "https://api.muapi.ai/api/v1"
headers = {"x-api-key": os.environ["MUAPI_API_KEY"]}

r = requests.post(f"{API}/gemini-3-8-flash-tts", headers=headers, json={"speakers":[{"speaker_id":"Speaker 1","voice_name":"Fenrir","audio_profile":"A stern and weary gatekeeper","accent":"British (RP)","style":"Deadpan","pace":"Natural"},{"speaker_id":"Speaker 2","voice_name":"Puck","audio_profile":"A determined and courageous traveler seeking answers.","accent":"American (Gen)","style":"Empathetic","pace":"Staccato"}],"dialogue_turns":[{"speaker_id":"Speaker 1","text":"[shouting] Halt, traveler! The northern pass is sealed by order of the council."},{"speaker_id":"Speaker 2","text":"[determination] I carry a message for the elder. Step aside, or I will force my way through."},{"speaker_id":"Speaker 1","text":"[caution] No one passes. [pensive] The elder is... he's no longer receiving visitors."},{"speaker_id":"Speaker 2","text":"It's too late. [whispers] The shadow... it reached him first. [urgency] You need to leave. [shouting] Now."}]})
request_id = r.json()["request_id"]

while True:
    res = requests.get(f"{API}/predictions/{request_id}/result", headers=headers).json()
    if res["status"] == "completed":
        print(res["outputs"]); break
    if res["status"] == "failed":
        raise RuntimeError(res.get("error")); 
    time.sleep(3)
```

## 输入结构

API 接受以下输入参数：

- **`speakers`** (`array`, _必填_):
  说话人声音配置列表。每一轮对话通过 ID 引用一个说话人。
  - **`speakers[].speaker_id`** (`string`, _必填_): 说话人标识符。必须使用 "Speaker N" 格式（例如 "Speaker 1"）。
  - **`speakers[].voice_name`** (`string`, _必填_): 预置 Gemini 声音名称。
    - 选项: `"Achernar"`, `"Achird"`, `"Algenib"`, `"Algieba"`, `"Alnilam"`, `"Aoede"`, `"Autonoe"`, `"Callirrhoe"`, `"Charon"`, `"Despina"`, `"Enceladus"`, `"Erinome"`, `"Fenrir"`, `"Gacrux"`, `"Iapetus"`, `"Kore"`, `"Laomedeia"`, `"Leda"`, `"Orus"`, `"Puck"`, `"Pulcherrima"`, `"Rasalgethi"`, `"Sadachbia"`, `"Sadaltager"`, `"Schedar"`, `"Sulafat"`, `"Umbriel"`, `"Vindemiatrix"`, `"Zephyr"`, `"Zubenelgenubi"`
  - **`speakers[].audio_profile`** (`string`, _可选_): 可选的 natural-language description of the persona, e.g. "A warm and soothing narrator".
  - **`speakers[].accent`** (`string`, _必填_): 说话口音。
    - 默认值: `"Neutral"`
    - 选项: `"Neutral"`, `"American (Gen)"`, `"American (Valley)"`, `"American (South)"`, `"British (RP)"`, `"British (Brixton)"`, `"Transatlantic"`, `"Australian"`
  - **`speakers[].style`** (`string`, _必填_): 情感表达风格。
    - 默认值: `"Empathetic"`
    - 选项: `"Vocal Smile"`, `"Newscaster"`, `"Whisper"`, `"Empathetic"`, `"Promo/Hype"`, `"Deadpan"`
  - **`speakers[].pace`** (`string`, _必填_): 说话速度。
    - 默认值: `"Natural"`
    - 选项: `"Natural"`, `"Rapid Fire"`, `"The Drift"`, `"Staccato"`

- **`dialogue_turns`** (`array`, _必填_):
  按顺序排列的对话行列表。每一轮的 speaker_id 必须匹配上面定义的说话人。文本可以包含 [shouting] 或 [whispers] 等语气标签。
  - **`dialogue_turns[].speaker_id`** (`string`, _必填_): 朗读此行的说话人 ID（例如 "Speaker 1"）。
  - **`dialogue_turns[].text`** (`string`, _必填_): 要朗读的台词。支持内嵌语气标签。最多 10000 个字符。

- **`scene`** (`string`, _可选_):
  可选的 scene description that sets the acoustic setting, e.g. "A quiet, warm room with a fireplace crackling softly."
  - 默认值: `""`

- **`sample_context`** (`string`, _可选_):
  可选的 overall tone/style, e.g. "Audiobook style narration. Tone is gentle and inviting."
  - 默认值: `""`

- **`temperature`** (`number`, _可选_):
  采样温度（0-2）。数值越高，表达方式越丰富。
  - 默认值: `1`
  - 范围: `0` to `2`

## 输出结构

轮询端点会返回以下字段：

- **`id`** (`string`): 请求 ID。
- **`status`** (`string`): 取值之一：`queued`、`pending`、`processing`、`completed`、`failed`、`cancelled`。
- **`outputs`** (`array`): 生成的图像、视频或音频 URL。`status` 为 `completed` 前为空。
- **`urls.get`** (`string`): 用于重新获取此预测结果的自身链接。
- **`error`** (`string` | `null`): `status` 为 `failed` 时的错误消息。
- **`created_at`** (`string`): 创建请求时的 ISO-8601 时间戳。
- **`has_nsfw_contents`** (`array of boolean`): 每个输出对应的 NSFW 检测标记。

## Webhook（可选）

在提交 URL 后附加 `?webhook=https://your-server/path`。任务进入终止状态后，MuApi 会将与轮询响应相同结构的 POST 请求发送到你的 URL，无需轮询。

## Agent 集成

MuApi 提供 MCP 服务器和 CLI，让 Agent（Claude Code、Cursor 或自定义 Agent）无需编写 HTTP 代码即可调用此端点：

```bash
# 安装 CLI
npm install -g muapi-cli

# 完成一次身份验证
muapi auth login

# 将所有 MuApi 模型作为 MCP 工具提供给 Agent
muapi mcp serve
```

MCP 服务器为每个模型提供封装提交与轮询的工具，包括 `gemini-3-8-flash-tts`。使用 `muapi --help` 查看按类别划分的快捷命令（`muapi image generate`、`muapi video from-image` 等）。

## 相关模型
- [3.1 Flash](https://muapi.ai/zh/playground/gemini-3-1-flash-tts)
- [2.5 Pro](https://muapi.ai/zh/playground/gemini-2-5-pro-tts)
- [3.8 Flash Lite](https://muapi.ai/zh/playground/gemini-3-8-flash-lite-tts)
- [3.8 Flash](https://muapi.ai/zh/playground/gemini-3-8-flash-tts)

## 资源
- [Playground 页面](https://muapi.ai/zh/playground/gemini-3-8-flash-tts)
- [API 文档](https://muapi.ai/zh/playground/gemini-3-8-flash-tts/api)
- [全站 llms.txt](https://muapi.ai/zh/llms.txt)